2021/03/19 08:50:42

Weak AI / strong AI and specialized / general-purpose type | AI / artificial intelligence terms

Weak AI / strong AI and specialized / general-purpose type | AI / artificial intelligence terms

Top Call Center Column AI
Strong AI / weak AI
● AI (artificial intelligence) classification method
There are several different methodologies for AI (artificial intelligence), and it is difficult to delineate whether or not it should be called AI (artificial intelligence) in the first place, but here, "specialized / general-purpose type" and "general-purpose type" and " I will explain the classification method of "weak AI / strong AI".

Strong AI / weak AI / general-purpose type / specialized type
● Specialized type and general-purpose type
There are two classification methods for AI (artificial intelligence): "specialized type" and "general-purpose type".

Specialized type is AI (artificial intelligence) that can exert its ability only in a specific specialized field. It specializes in AI (artificial intelligence) that allows you to play Go, Shogi, and chess, as well as specialized fields such as law, medical care, and failure diagnosis. It can be said that the expert system of a generation ago is also a specialized AI, but at that time it was necessary for humans to teach (describe) the rules, so the technology used is the technology that is popular now that utilizes deep learning. It's different.

On the other hand, the general-purpose type is one that can make inferences for a wide range of objects like humans. The humanoid androids and anime Doraemon that appear in science fiction movies can be said to be robots equipped with general-purpose AI (artificial intelligence).

In order to realize AI (artificial intelligence) that can respond to anything like human beings, it is thought that AI (artificial intelligence) itself needs the ability to learn and evolve. This is because there is a limit to how humans can code everything and teach it to AI (artificial intelligence).

● What is weak AI / strong AI?
There is a classification method of weak AI and strong AI, which is different from the general-purpose type and specialized type.

Rule-based reasoning (artificial intelligence), in which humans a generation ago describe rules, is a weak AI. We cannot surpass humans because we cannot do more than what humans have taught us, such as law and medical care. Go battle software is also a specialized "weak AI" if it is as good as a child's playmate.

On the contrary, strong AI is AI (artificial intelligence) that has better reasoning ability than human beings. AlphaGo was as strong as even a professional top-level Go player. It can be said to be a specialized "strong AI".

Deep learning technology is at the base of strong AI. And we need many examples for learning. In addition, you may want to take advantage of modern world's fastest supercomputers, such as AlphaGo. However, the computing power of computers will evolve over time, so one day we should be able to perform arithmetic processing comparable to AlphaGo on personal computers.

And in specialized AI such as finding similar cases in trials and finding lesions from X-ray photographs, "strong AI" that has already exceeded human ability has emerged. However, weak AI does not mean that weak AI is completely useless, and weak AI may also have business opportunities.

By the way, general-purpose AI should have the ability to learn by itself, so even if it is a "weak AI" at first, it should eventually grow into a "strong AI". There is a condition that there is no problem with the computing power of the computer. And, some kind of breakthrough is necessary for such "general-purpose AI" to be realized, and at this point it may not be clear whether it can be realized yet.